Main

Follicular dendritic cell sarcoma is a rare neoplasm of mesenchymal lineage that was first described by Rosai and colleagues.1 Follicular dendritic cell sarcoma pursues a variable and largely unpredictable clinical course. Some patients (especially those with isolated nodal involvement) are cured by surgical excision, whereas others develop locally recurrent or metastatic disease. No reliable predictors of disease course have been identified.2, 3 In addition, little is known about the genetic alterations that drive tumorigenesis in follicular dendritic cell sarcoma. Although karyotypic abnormalities have been cited in case reports, no recurrent alterations have been identified.4, 5 To our knowledge, the only study to investigate the genetic basis of follicular dendritic cell sarcoma was recently reported by Go et al,6 who looked for alterations in BRAF and identified V600E mutations in 5 of 27 cases (19%). The aim of the present study was to identify somatic genetic alterations in follicular dendritic cell sarcoma using a targeted next-generation sequencing assay that detects alterations in 309 known cancer-related genes. We evaluated 13 cases of follicular dendritic cell sarcoma reviewed at our institution and found recurrent alterations in genes involved in the regulation of NF-κB activation (NFKBIA and CYLD), cell cycle progression (CDKN2A and RB1), and immune evasion (CD274 and PDCD1LG2). These findings represent novel insights into the genomic landscape of follicular dendritic cell sarcoma and highlight commonalities in tumorigenesis with certain other tumor types, notably subsets of B-cell lymphoma.

Materials and Methods

Case Identification

An institutional review board-approved search of the pathology archives at Brigham and Women’s Hospital from 2004 to 2015 yielded 13 cases of follicular dendritic cell sarcoma with sufficient formalin-fixed paraffin-embedded material for sequencing. The diagnosis of follicular dendritic cell sarcoma was confirmed by review of H&E slides and available immunohistochemistry. All cases were conventional follicular dendritic cell sarcomas, including the primary hepatic tumors, which did not show histological features of the inflammatory pseudotumor-like variant. By immunohistochemistry, all tumors were positive for CD21, and 12 of 12 evaluated cases were positive for CD35. In addition, 5 of 6 cases were positive for podoplanin (D2-40), and 3 of 10 were focally positive for epithelial membrane antigen. All studied tumors were negative for S-100 protein and keratins (5 and 7 cases, respectively). Additional information including patient demographics and clinical follow-up was obtained from the medical record when available.

Targeted Next-Generation Sequencing Assay and Data Processing

The tumor cellularity in the samples ranged from 50 to 80%. Targeted sequencing of 309 cancer-related genes (see Supplementary Table 1 for a complete list) was performed on DNA isolated from formalin-fixed paraffin-embedded tumor samples using hybrid capture with a custom probe set (Agilent SureSelect, Agilent Technologies, Santa Clara, CA) and massively parallel sequencing (HiSeq 2500, Illumina, San Diego, CA), as previously described.7 Data processing and analysis were performed using MuTect and GATK for single-nucleotide variants,8, 9 VisCap Cancer for copy-number variations (Dana Farber Cancer Institute, Boston, MA), and BreakMer for large structural variations.10 Manual review and annotation were performed for all single-nucleotide variants, copy-number variations, and translocation calls made by the data pipeline. All single-nucleotide variants with a minor allele frequency >0.1% in the Exome Sequencing Project database were excluded as likely single-nucleotide polymorphisms. For single-nucleotide variants reported in the Exome Sequencing Project with a minor allele frequency of <0.1%, further analysis with Mutation Assessor (Memorial Sloan Kettering Cancer Center, New York, NY) and PolyPhen2 (ref. 11) was performed to assess the predicted effect on protein function, with single-nucleotide variants retained only if they were classified as deleterious by both platforms. All retained alterations were then classified as likely pathogenic variants or variants of unknown significance according to the following criteria. Likely pathogenic variants were defined as frameshift, nonsense, or splice-site mutations; known pathogenic missense alterations; bi-allelic losses in tumor suppressors; or focal copy-number gains in oncogenes. Variants of unknown significance were defined as missense mutations with unknown functional effects or low-level copy-number gains and losses of uncertain function.

Immunohistochemistry and In Situ Hybridization

Immunohistochemistry was performed on selected cases following pressure cooker antigen retrieval (Target Retrieval Solution, pH 6.1; Dako, Carpinteria, CA) using a mouse anti-p16 monoclonal antibody (clone E6H4; 1:2 dilution; Ventana, Tucson, AZ) and a mouse anti-RB1 monoclonal antibody (clone G3-245; 1:100 dilution; BD Biosciences, San Diego, CA). In situ hybridization studies for Epstein–Barr virus-encoded mRNA (EBER) were performed on a Ventana Benchmark XT autostainer using an EBER 1 DNP probe (Ventana, cat # 760-1209A) and ISH iView Blue Plus anti-DNP detection system (Ventana).

Results

Clinical characteristics (including available follow-up information) for each case are summarized in Table 1. The mean patient age was 48 years (range 26–70 years). Eight patients were male and five were female. Nine tumors involved extranodal sites (liver–3, chest wall–1, omentum–1, bile duct–1, pre-sacral retroperitoneum–1, ovary–1, and supraclavicular fossa–1) and four tumors involved lymph nodes (cervical–2, axillary–1, and tonsil and neck–1). The predominantly extranodal distribution of disease in this cohort (as opposed to the lymph node predominant distribution typical of follicular dendritic cell sarcoma) is likely a reflection of the subspecialty expertise in soft tissue tumor pathology of the senior authors of this study. Eight cases were primary diagnoses without prior treatment, one represented residual disease after chemoradiation, one was a local recurrence after surgery, and three were metastases following surgery, chemotherapy, or radiation. Clinical follow-up data were available for six cases and included one patient dead from disease 37 months after primary diagnosis (case 4), three patients alive with disease 17, 56, and 12 months after primary diagnosis (cases 2, 5, and 13, respectively), and two patients alive with no evidence of disease 61 and 27 months after primary diagnosis (cases 1 and 3, respectively). All 12 cases evaluated by in situ hybridization for EBER (including the two primary hepatic tumors) were negative. Sequencing quality for all samples exceeded the minimum benchmarks established in our laboratory, with a mean total coverage of 209 (range 52–310) and a mean percentage of sequences achieving greater than 30-fold coverage of 97% (range 83–99%). The copy-number variations data for one case (case 6) were not suitable for interpretation due to signal dispersion, which was likely caused by DNA degradation, and were excluded. Genes that included at least one likely pathogenic variant are summarized in Figure 1 along with any concurrent variants of unknown significance. All remaining alterations, including genes with variants of unknown significance but no likely pathogenic variants, are included in Supplementary Table 2.

Table 1 Case characteristics
Figure 1
figure 1

Summary of likely pathogenic alterations identified in 13 cases of follicular dendritic cell sarcoma. Included in the figure are all genes for which at least one likely pathogenic variant was observed. For each of these genes, all concurrent variants of unknown significance are also included. Genes with recurrent variants of unknown significance but no likely pathogenic variants are included in Supplementary Table 2. The use of two different colors for a particular gene indicates that two separate alterations were observed. *No copy-number variation data were obtained for case 6 due to signal dispersion likely caused by DNA degradation.

Tumor suppressor genes critical for the negative regulation of NF-κB activation showed recurrent likely pathogenic variants in 5 of 13 (38%) cases of follicular dendritic cell sarcoma (Figures 1 and 3). These alterations occurred in a mutually exclusive manner and included three cases with frameshift mutations or deletions in NFKBIA (cases 1, 11, and 12), which encodes a protein involved in the cytoplasmic sequestration of the NF-κB complex. In addition, two cases demonstrated bi-allelic loss of CYLD (cases 2 and 4), which encodes a protein that regulates NF-κB activation through the deubiquitination of upstream mediators. Of note, both cases with loss of CYLD were metastases and carried concurrent likely pathogenic variants in at least one cell cycle regulatory gene (CDKN2A and/or RB1, as described below). In addition, three cases had variants of unknown significance in NFKBIA, including one case with a missense mutation (N108K, case 13) and two cases with one-copy losses (cases 7 and 10). One case also had a previously reported missense mutation in TNFAIP3 (N102S, case 4), which encodes a protein that (similar to CYLD) regulates NF-κB activation via deubiquitination of upstream mediators.12 Interestingly, this variant occurs in the deubiquitinating domain of TNFAIP3 and has been reported both in dbSNP (rs146534657) with a minor allele frequency of 1.3% and in several cases of B-cell lymphoma.12 The observed allele fraction for this variant (60%) in case 4 could be consistent either with a single-nucleotide polymorphism or a somatic change demonstrating copy neutral loss of heterozygosity. Finally, low-level copy-number gain on chromosome 20q11 was noted in six cases in a region containing the anti-apoptotic gene BCL2L1, which is a primary transcriptional target of NF-κB.13

Cell cycle regulatory genes also showed recurrent alterations in follicular dendritic cell sarcoma, with likely pathogenic variants in the canonical tumor suppressors CDKN2A or RB1 noted in 4 of 13 (31%) cases (Figure 1). These included two cases with bi-allelic loss of CDKN2A (cases 2 and 4), one case with bi-allelic loss of RB1 (case 4), and one case with a nonsense mutation in RB1 (case 6). Each of these cases showed loss of p16 and Rb protein expression, respectively, by immunohistochemistry (Figure 2). Additional notable alterations in cell cycle regulatory genes included a focal MYC copy-number gain in one case (case 12); copy-number gains involving CCND2 (six cases), CDK4 and MDM2 (four cases), and KRAS (four cases); and one-copy losses in CDKN2A (four cases), RB1 (two cases), and TP53 (two cases). Although the functional consequences of the one-copy losses observed in tumor suppressors is uncertain due to the apparent absence of a second ‘hit’, alternative mechanisms of gene silencing (e.g., methylation) are not detected by the sequencing assay and could potentially contribute to bi-allelic inactivation in these cases. Additional notable alterations included three cases (cases 3, 4, and 8) with focal copy-number gains on chromosome 9p24 in a region containing the immune-regulatory genes CD274 (PD-L1), PDCD1LG2 (PD-L2), and JAK2. Pertinent negative findings include the absence of any BRAF V600E mutations in the cohort, which contrasts with the findings of Go et al6 cited above. Finally, translocation analysis did not identify any recurrent or previously described translocations. A novel t(5;9)(p15;q22) involving SYK and EXOC3 was noted in one case (case 2). However, as this translocation has not been reported previously and was only present in one case, its significance is uncertain.

Figure 2
figure 2figure 2

P16 and RB expression in follicular dendritic cell sarcoma. (a, c, e and g) Representative H&E images from case 1 (a), case 4 (c), case 2 (e), and case 6 (g). (b and d) Representative images of immunohistochemistry for P16 for case 1 (b) and case 4 (d). (f and h) Representative images of immunohistochemistry for RB for case 2 (f) and case 6 (h). All images original magnification x400. H&E, hematoxylin and eosin.

Discussion

In the present study, likely pathogenic loss-of-function alterations in NF-κB regulatory genes occurred in 5 of 13 (38%) cases. These alterations occurred in a mutually exclusive manner, which is consistent with their functioning in the same pathway. In addition, loss-of-function alterations in NF-κB and cell cycle regulatory genes co-occurred in two of the three metastatic tumors included in the study (cases 2 and 4), suggesting that the accumulation of multiple pathogenic mutations in these pathways may possibly correlate with a more aggressive disease course. Notably, among the six patients for whom clinical follow-up information was available, one patient (case 4) had both the highest apparent mutational burden (involving four concurrent likely pathogenic variants, which included loss-of-function alterations in CYLD, CDKN2A, and RB1, as well as copy-number gain of 9p24) and the most aggressive disease course (dying from metastatic disease 37 months after initial diagnosis). Although a larger study would be required to confirm any association between molecular features and clinical behavior, these findings suggest that molecular analysis could potentially provide a useful adjunct for the prediction of disease course in patients with follicular dendritic cell sarcoma.

Although mutations in cell cycle regulatory genes are frequently found across tumor types, loss-of-function alterations in NF-κB regulatory genes, including NFKBIA, CYLD, and TNFAIP3, are less common but have been observed most frequently in lymphoid neoplasms, including Hodgkin lymphoma,14 diffuse large B-cell lymphoma (Figure 3),15, 16 and plasma cell myeloma,17 as well as select solid tumors, including adenoid cystic carcinoma18 and metastatic prostatic adenocarcinoma (Figure 3).19 NFKBIA, CYLD, and TNFAIP3 are thought to exert tumor suppressor function primarily through the negative regulation of the canonical pathway of NF-κB activation. Under normal conditions, NF-κB activation through this pathway involves several key steps, including: (1) activation of the IκB kinase in response to an appropriate extracellular stimulus; (2) phosphorylation, ubiquitination, and proteosomal degradation of IκB-alpha (encoded by NFKBIA), which otherwise binds to NF-κB in the cytoplasm and inhibits its nuclear translocation; (3) release and nuclear translocation of NF-κB (itself composed of heterodimers of the proteins p50 or p52 and REL, RELA, or RELB); and (4) expression of NF-κB target genes.13 Although IκB-alpha regulates NF-κB via direct interactions, the proteins encoded by both CYLD (CYLD) and TNFAIP3 (A20) exert indirect regulatory activity through the removal of ubiquitin chains from upstream signaling molecules, such as RIP1 and TRAF2, which are important for activation of the IκB kinase.20, 21

Figure 3
figure 3

Alterations in the NF-κB regulatory pathway in follicular dendritic cell sarcoma and other tumor types. (a) Gene level view of NFKBIA showing the location of each observed mutation. (b) Chromosome level view of chromosome 16 for cases 2 and 4 demonstrating bi-allelic loss of CYLD, as well as flanking one-copy loss of RBL1. (c) Frequency of loss-of-function alterations (bi-allelic loss, frameshift mutation, and nonsense mutation) in the NF-κB regulatory genes NFKBIA, CYLD, and TNFAIP3 across cancer genome studies available in cBioportal, which was used to generate the figure.15 Insets show frequency of alterations in each gene from selected studies. Letters and numbers listed in parentheses indicate the name of the corresponding cell line. ACyC, adenoid cystic carcinoma; BCL, B-cell lymphoma; CCLE, Cancer Cell Line Encyclopedia; CS, carcinosarcoma; DLBC, diffuse large B-cell lymphoma; HL, Hodgkin lymphoma; lung squ, lung squamous cell carcinoma; MM, multiple myeloma; MSKCC, Memorial Sloan Kettering Cancer Center; MICH, University of Michigan; TCGA, The Cancer Genome Atlas.

In the setting of cancer, loss-of-function alterations in NFKBIA, CYLD, or TNFAIP3 lead to increased NF-κB nuclear translocation and increased expression of NF-κB target genes important for cell proliferation and cell survival.13, 22 Foremost among these targets are the inhibitor of apoptosis and a subset of BCL2-family genes, including BCL2L1, which exert anti-apoptotic activity through the inhibition of caspases and pro-apoptotic BCL2-family genes, respectively.23 Interestingly, the most frequent copy-number variation observed in the study was low copy-number gain on 20q11 in a region that includes BCL2L1, as well as several other candidate genes (observed in six cases). Although the impact of this alteration in follicular dendritic cell sarcoma is uncertain, a functional role of BCL2L1 activation and copy-number gain has been demonstrated in other tumor types24 and may suggest future avenues of investigation into potential mediators of anti-apoptotic activity in follicular dendritic cell sarcoma, as well as potential therapeutic targets, given recent interest in the development of inhibitors of BCL-XL, the protein encoded by BCL2L1.25

Finally, copy-number gain on chromosome 9p24 was observed in three cases in a region containing the genes CD274 (PD-L1) and PDCD1LG2 (PD-L2). Copy-number gain on 9p24 represents a well-described mechanism of immune evasion in several subtypes of lymphoma, including nodular sclerosing Hodgkin lymphoma and primary mediastinal diffuse large B-cell lymphoma, and is mediated in part by increased tumor expression of PD-L1 and PD-L2 and consequent activation of the PD-1 regulatory pathway on tumor-infiltrating lymphocytes.26, 27, 28 Although further study would be required to clarify the specific role of immune evasion in the pathogenesis of follicular dendritic cell sarcoma, the finding of 9p24 copy-number gain in a subset of cases raises the possibility that immunotherapy with PD-1/PD-L1/2 pathway blockade may represent a future avenue of therapeutic investigation.27 In summary, the current findings represent novel insights into the genomic landscape of follicular dendritic cell sarcoma and provide a foundation for further experimental and therapeutic investigation in this tumor type.